Cost Reduction in Ultrahigh-Purity Gas Delivery Systems

In the fabrication of high-purity tubing (Fig. 1), the prevailing thought has been that a narrower weld provides greater resistance to corrosion inside a stainless-steel tubing system. For example, a narrow 1T weld bead has a root thickness of 1× the thickness of the tubing wall itself, while a wider 2T weld bead has a root thickness 2× the wall thickness (Fig. 2). In the case of a 3/8 in. tube (outer diameter, or OD) and a wall thickness of 0.035 in., the 1T root width is 0.035±0.009 in. and the 2T root width is 0.07±0.018 in. Such is the case across all common tubing diameters including 1/4, 3/8, 1/2, 1 and 2 in. OD (Table 1).

The practice of using 1T bead widths as a standard in much of the industry is based on the fact that corrosion has a tendency to start in and around the weld bead and the heat-affected zones (HAZ) on each side of the bead. The prevailing opinion has been that, by minimizing the size of the weld bead and heat input used to complete the weld, the HAZ on either side of the weld should be smaller. Therefore, by routinely using narrower weld beads, the susceptibility to corrosion should potentially be greatly reduced.

We have recently performed extensive tests of weld beads in stainless-steel tubing. We discovered that using a 1T bead width does not provide better corrosion resistance than a 2T bead width. In fact, there was no measurable difference in pitting corrosion resistance between the two bead widths.

These findings not only refute earlier assertions that 1T bead widths improve the overall corrosion resistance of gas delivery systems, but they also strongly indicate that significant savings of up to 30% of system cost could be realized if the industry began to specify 2T instead of 1T bead widths. In addition, we found that the quality and consistency of weld beads is significantly better for 2T systems.

	1. The process of fabricating stainless-steel tubing with 1T weld bead widths can cost up to a third more than fabricating 2T width tubing for the same system, without providing any additional benefit of corrosion resistance.

The industry itself has estimated that the specification of 1T weld beads conservatively adds as much as 30% to the total welding costs associated with system fabrication. This can amount to millions of dollars when extensive and complex gas delivery systems are involved. Weld beads of 1T width also are more susceptible to lack of penetration and other defects, such as bead meandering. Although bead meandering almost always is a factor in welding, the potential problems that it can create increase significantly with narrower weld beads.

Despite these issues, 1T bead widths have been an accepted practice because of the belief that the extra costs will bring long-term benefits through enhanced performance and corrosion resistance. But new data clearly indicate that using 2T bead widths offers additional benefits. First, 2T bead widths can significantly reduce the costs of fabricating and retrofitting systems because it is easier to obtain an acceptable weld. Weld operators can spend more time producing product and less time qualifying and couponing weld procedures. Secondly, 2T bead widths ensure against lack of penetration. The wider 2T weld bead compensates for bead meandering, making it more likely each weld will meet requirements. Finally, the fact that corrosion resistance between 1T and 2T bead widths is comparable indicates that the problem of corrosion has more to do with the composition of the stainless-steel tubing than with welding. This suggests that system designers should focus on material selection rather than bead width and heat input to minimize susceptibility to corrosion.

	2. This schematic shows the relationship between tubing wall thickness and the weld bead width, which is available as 1T or 2T.

As gas delivery systems came to be designed and built on smaller scales, researchers shifted their attention to the number of weld joints and width of weld beads. Since weld joints comprised a larger percentage of the tubing wall surface, they presented potential weak links for corrosion and leakage in the system.

Early research from the 1980s prompted an industry trend to 1T weld beads in highly corrosive gas delivery systems.^1,2 Today, more than half the system designers in the industry regularly specify 1T weld beads.

In making the case for 1T welds, researchers examined the passivated and unpassivated areas of the tubing wall. The interior, electropolished surface of the tubing is passivated, and each weld introduces unpassivated surface area. Early research attempted to link this unpassivated surface to corrosion problems. The presence of contamination downstream from the welds fueled speculation that the likely sources were the unpassivated areas, where the potential for pitting and entrapment is greatly increased.

The research then focused on ways to reduce the unpassivated areas — in particular, to minimize bead width — as a means of enhancing corrosion resistance. Under this rationale, a 1T weld of 0.035 in. in a system with 200 welds would result in 7 in. of unpassivated surface. If 2T bead widths are used, the unpassivated surface doubles to 14 in.

But until this study by Swagelok, the actual risk of greater corrosion in 2T weld beads, relative to 1T weld beads, was never rigorously evaluated. Because a greater heat input is required to make a 2T weld, a 2T weld bead differs in penetration, heat input and mechanical characteristics.

But the effect of these differences on corrosion resistance was not known. We do know that these factors, taken to extreme, will cause other quality problems. For example, the dramatically increased heat input required to make 3T welds in an electropolished 316L tube system causes mechanical problems, including bead sagging, making them unacceptable for use.

Through its involvement in the development of advanced orbital welding systems and techniques, Swagelok has conducted extensive research into the many complex issues surrounding weld quality and stability. One research area centered on evaluating bead width and corrosion response. Because original research in this area was based on only a few specimens, we used an expanded test sample and matrix.

We replicated earlier testing with a test sample of 1T and 2T welds over a wide range of sizes and heats to establish a baseline for the industry. We used ASTM G 150,³ "Standard Test Method for Electrochemical Critical Pitting Temperature (CPT) Testing of Stainless Steel," for the basis of comparison in the study. (Note that SEMI's Corrosion Task Force is charged with developing a standardized corrosion test method that compares the performance of materials in corrosive gases used by the semiconductor industry, based on ASTM G 150.)

We performed welding and corrosion tests on specimens from 18 heats of commercially available tubing, representing a variety of mills, composition, sizes, and melting and refining processes. The 316L tubing compositions ranged from high nickel and high chromium content to low manganese content. Samples from each tube lot were welded using the company's M-100 welding system. This equipment permits data to be gathered on every weld performed, including information on current, voltage and travel speed, from which heat inputs can be calculated. Welding parameters were varied to achieve an acceptable weld, as defined by weld bead thickness and appearance. We fabricated two sets of samples from each heat, with 1T and 2T weld beads. We noted the number of setups necessary to achieve a reproducible weld schedule.

The following summarizes the testing, comparing the CPT results of 1T vs. 2T weld beads:

• 18 heats of tubing were tested, with six welds performed on each heat.
• On a heat-by-heat basis, in 17 out of 18 cases, the sample means, standard deviations and variances were not different at a confidence level of 97.5%.
• Comparing 1T averages vs. 2T averages, the population means, standard deviations and variances were not different at a confidence level of 97.5%.

The results indicate that cost savings can be achieved through the greater use of 2T weld beads. Throughout this study, we tracked the number of coupons — or weld setups — it took to create a weld schedule that would produce acceptable 1T and 2T welds, using an advanced, computerized welding system. It took 3× as many setups to achieve quality 1T welds as 2T welds.

Considering it can take from 10 to 30 minutes to develop each coupon, the amount of time that can be saved from switching to 2T weld beads is staggering. The study found that it took between two and 29 weld setups to achieve acceptable 1T welds, while the setups required for quality 2T welds ranged from two to 15, with identical equipment and operator skills. Even the most skilled weld operators can struggle with 1T setups.

Reducing the weld bead from 2T to 1T also has a significant effect on weld quality, as evidenced by the percentage of rewelds required to achieve the narrower bead width. Inspection of the bead widths with bore scopes and site pipes found that, as the bead width narrowed, the chances of having slight variations in quality increased greatly.

Our tests provide solid support for the use of 2T bead widths in gas delivery systems, both from a quality and cost-savings standpoint. We showed there is no statistically measurable difference between the corrosion resistance of 1T and 2T weld beads made in identical 316L tubing samples over a range of 316L material compositions.

This is of particular importance to the semiconductor industry, where high performance is paramount and capital investment in new systems, retrofits and upgrades is undertaken on an ongoing basis. The sheer number of weld joints in a given system necessitates superior consistency and quality in every connection. Clearly, 2T bead widths will meet these requirements with less setup and rework while offering corrosion and mechanical performance equivalent to 1T welds.

We then tested another 18 heats of commercially available 316L electropolished tubing. Again, we evaluated the welded tube samples from each lot for corrosion resistance using standard ASTM test methods. Welding parameters were chosen to achieve an acceptable weld, as defined by weld bead thickness and appearance. ASTM G 150 critical pitting temperature tests were performed. Only the internal electropolished surface of the tubing was exposed. More than 300 individual corrosion tests were performed to obtain a body of reproducible data.

Table 2 shows typical welding parameters for each heat of tubing. The current was established to achieve the bead width of 1T or 2T. The welding equipment sets the voltage for the current and the arc gap. We performed the welding using ceriated tungsten electrodes, and shield and purge gas of 100% argon. The study found it requires ~15% greater heat input needed to create a 2T weld bead. However, this amount of additional heat input had no measurable effect on 2T weld performance or corrosion resistance.

The compositions of the 18 heats are given in Table 3, along with the tubing sizes of each heat. These values were determined against traceable 316 SS standards, by X-ray fluorescence (Cr, Mn, Mo, Ni, Si), optical emission spectroscopy (P, Cu), and high-temperature combustion/inert gas fusion (C, S, N).

Each as-received specimen of tubing was 2 in. long. To create a welded specimen, two 2 in. lengths of identical material were orbital/autogenously welded into 4 in. segments. We cut the welded specimens so that the weld was 0.5 in. from one end, and the total length was 2 in. We masked all specimens with lacquer and Plater's tape to isolate the test surface to the inner surface only. About 1 in. of the specimen was submerged vertically into solution with the weld 0.5 in. from the submerged tube end.

The test cell was a 2 L glass beaker cut to 4.5 in. high and capped with a polyethylene cover. We used a three-electrode setup consisting of the test sample, an SCE reference electrode inserted in a Luggin probe, and two graphite rods as counter electrodes. All testing was performed with Gamry Instruments software (CMS100) and potentiostats (PC3/750).

The test solution was 3.56% NaCl made from reagent-grade chemicals and DI (17 MΩ) water. On completion of the test, the specimen was removed from the cell and rinsed in DI water. Each specimen was examined to identify the location of the pitting. Specimens that showed signs of corrosion due to insufficient lacquer were discarded. We performed at least six valid repetitions for each heat.

We performed the electrochemical CPT test (ASTM G 150) to determine resistance to localized (pitting) corrosion. The CPT test measures the temperature at which current density increases rapidly beyond a set limit at a set electrical potential. The electrical potential is held constant in the passive region. Starting at 0°C, the temperature is raised slowly until pitting occurs. Higher CPT signifies greater resistance to pitting corrosion.

We allowed each specimen to equilibrate for 10 min at 0°C. A 700 mV SCE potential was applied to the specimen while the temperature was raised at a rate of 1°C/min, measuring current continuously. The test ended when the current density exceeded 100 µA/cm² for 60 sec, signifying the onset of localized corrosion.

Table 4 lists the results for as-received and welded specimens. As anticipated, welding generally reduces the corrosion resistance of the test materials. The table reports the descriptive statistics of samples consisting of a minimum of six repeats for each condition.

We performed a parametric evaluation of the differences between the two groups of data.⁴ In 17 of the 18 heats, we calculated no statistically significant difference between the sample standard at a 97.5% level. Only one heat showed sample standard deviations that were significantly different.

Comparing 1T averages with 2T averages, the population standard deviations and the population means were not different at a confidence level of 97.5%.

These results show that there is no advantage gained in assembly, corrosion resistance, performance or cost by minimizing weld bead widths to 1T from 2T, in orbitally welded 316L electropolished tubing.

Based on extensive welding data, the amount of heat input necessary to go from a 1T to a 2T weld averages ~15%. This additional heat input makes fabrication of reproducible and consistent welds much easier, as evidenced by the reduction in the number of setups. The study confirms practical experience, demonstrating that 1T welds are more difficult and more expensive to set up and produce consistently than 2T welds. On average, producing reproducible 1T welds requires 3× the setup time of 2T welds.

In addition, the increase in heat input for the 2T welds was not sufficient to create a difference in corrosion response of the welds. Using CPT measurements, we found no statistically significant difference in pitting corrosion resistance between 1T and 2T welds.